Improving the Performance of Hyperspectral Target Detection PDF Download
Are you looking for read ebook online? Search for your book and save it on your Kindle device, PC, phones or tablets. Download Improving the Performance of Hyperspectral Target Detection PDF full book. Access full book title Improving the Performance of Hyperspectral Target Detection by . Download full books in PDF and EPUB format.
Author:
Publisher:
ISBN:
Category : Dimension reduction (Statistics)
Languages : en
Pages :
Get Book Here
Book Description
This dissertation develops new approaches for improving the performance of hyperspectral target detection. Different aspects of hyperspectral target detection are reviewed and studied to effectively distinguish target features from background interference. The contributions of this dissertation are detailed as follows. 1) Propose an adaptive background characterization method that integrates region segmentation with target detection. In the experiments, not only unstructured matched filter based detectors are considered, but also two hybrid detectors combining fully constrained least squared abundance estimation with statistic test (i.e., adaptive matched subspace detector and adaptive cosine/coherent detector) are investigated. The experimental results demonstrate that using local adaptive background characterization, background clutters can be better suppressed than the original algorithms with global characterization. 2) Propose a new approach to estimate abundance fractions based on the linear spectral mixture model for hybrid structured and unstructured detectors. The new approach utilizes the sparseness constraint to estimate abundance fractions, and achieves better performance than the popular non-negative and fully constrained methods in the situations when background endmember spectra are not accurately acquired or estimated, which is very common in practical applications. To improve the dictionary incoherence, the use of band selection is proposed to improve the sparseness constrained linear unmixing. 3) Propose random projection based dimensionality reduction and decision fusion approach for detection improvement. Such a data independent dimensionality reduction process has very low computational cost, and it is capable of preserving the original data structure. Target detection can be robustly improved by decision fusion of multiple runs of random projection. A graphics processing unit (GPU) parallel implementation scheme is developed to expedite the overall process. 4) Propose nonlinear dimensionality reduction approaches for target detection. Auto-associative neural network-based Nonlinear Principal Component Analysis (NLPCA) and Kernel Principal Component Analysis (KPCA) are applied to the original data to extract principal components as features for target detection. The results show that NLPCA and KPCA can efficiently suppress trivial spectral variations, and perform better than the traditional linear version of PCA in target detection. Their performance may be even better than the directly kernelized detectors.
Author:
Publisher:
ISBN:
Category : Dimension reduction (Statistics)
Languages : en
Pages :
Get Book Here
Book Description
This dissertation develops new approaches for improving the performance of hyperspectral target detection. Different aspects of hyperspectral target detection are reviewed and studied to effectively distinguish target features from background interference. The contributions of this dissertation are detailed as follows. 1) Propose an adaptive background characterization method that integrates region segmentation with target detection. In the experiments, not only unstructured matched filter based detectors are considered, but also two hybrid detectors combining fully constrained least squared abundance estimation with statistic test (i.e., adaptive matched subspace detector and adaptive cosine/coherent detector) are investigated. The experimental results demonstrate that using local adaptive background characterization, background clutters can be better suppressed than the original algorithms with global characterization. 2) Propose a new approach to estimate abundance fractions based on the linear spectral mixture model for hybrid structured and unstructured detectors. The new approach utilizes the sparseness constraint to estimate abundance fractions, and achieves better performance than the popular non-negative and fully constrained methods in the situations when background endmember spectra are not accurately acquired or estimated, which is very common in practical applications. To improve the dictionary incoherence, the use of band selection is proposed to improve the sparseness constrained linear unmixing. 3) Propose random projection based dimensionality reduction and decision fusion approach for detection improvement. Such a data independent dimensionality reduction process has very low computational cost, and it is capable of preserving the original data structure. Target detection can be robustly improved by decision fusion of multiple runs of random projection. A graphics processing unit (GPU) parallel implementation scheme is developed to expedite the overall process. 4) Propose nonlinear dimensionality reduction approaches for target detection. Auto-associative neural network-based Nonlinear Principal Component Analysis (NLPCA) and Kernel Principal Component Analysis (KPCA) are applied to the original data to extract principal components as features for target detection. The results show that NLPCA and KPCA can efficiently suppress trivial spectral variations, and perform better than the traditional linear version of PCA in target detection. Their performance may be even better than the directly kernelized detectors.
Author: Shlomi Hai Bouganim
Publisher:
ISBN:
Category : Algorithms
Languages : en
Pages : 80
Get Book Here
Book Description
Author: Carl Joseph Cusumano
Publisher:
ISBN:
Category :
Languages : en
Pages : 49
Get Book Here
Book Description
Hyperspectral imaging sensors suffer from pixel-to-pixel response nonuniformity that manifests as fixed pattern noise (FPN) in collected data. FPN is typically removed by application of flat-field calibration procedures and nonuniformity correction algorithms. Despite application of these techniques, some amount of residual fixed pattern noise (RFPN) may persist in the data, negatively impacting target detection performance. In this work we examine the conditions under which RFPN can impact detection performance using data collected in the SWIR across a range of target materials. We designed and conducted a unique tower-based experiment where we carefully selected target materials that have varying degrees of separability from natural grass backgrounds. Furthermore, we designed specially-shaped targets for this experiment that introduce controlled levels of mixing be tween the target and background materials to support generation of high fidelity receiver operating characteristic (ROC) curves in our detection analysis. We perform several studies using this collected data. First, we assess the detection performance after a conventional nonuniformity correction. We then apply several scene-based nonuniformity correction (SBNUC) algorithms from the literature and assess their abilities to improve target detection performance as a function of material separability. Then, we introduced controlled RFPN and study its adverse affects on target detection performance as well as the SBNUC techniques' ability to remove it. We demonstrate how residual fixed pattern noise affects the detectability of each target class differently based upon its inherent separability from the background. A moderate inherently separable material from the background is affected the most by the inclusion of SBNUC algorithms.
Author: Adam Cisz
Publisher:
ISBN:
Category : Computer algorithms
Languages : en
Pages : 262
Get Book Here
Book Description
"This thesis performs a performance comparison on existing hyperspectral target detection algorithms. The algorithms chosen for this analysis include multiple adaptive matched filters and the physics based modeling invariant technique. The adaptive matched filter algorithms can be divided into either structured (geometrical) or unstructured (statistical) algorithms. The difference between these two categories is in the manner in which the background is characterized. The target detection procedure includes multiple pre-processing steps that are examined here as well. The effects of atmospheric compensation, dimensionality reduction, background characterization, and target subspace creation are all analyzed in terms of target detection performance. At each step of the process, techniques were chosen that consistently improved target detection performance. The best case scenario for each algorithm is used in the final comparison of performance. The results for multiple targets were computed and statistical matched filter algorithms were shown to outperform all others in a fair comparison. This fair comparison utilized a FLAASH atmospheric compensation for the matched filters that was equivalent to the physics based invariant process. The invariant technique was shown to outperform the geometric matched filters that it uses in its approach. Each of these techniques showed improvement over the SAM algorithm for three of the four targets analyzed. Multiple theories are proposed to explain the anomalous results for the most difficult target"--Abstract.
Author:
Publisher:
ISBN:
Category :
Languages : en
Pages : 389
Get Book Here
Book Description
This research extends the field of hyperspectral target detection by developing autonomous anomaly detection and signature matching methodologies that reduce false alarms relative to existing benchmark detectors. The proposed anomaly detection methodology adapts multivariate outlier detection algorithms for use with hyperspectral datasets containing thousands of high-dimensional spectral signatures. In so doing, the limitations of existing, non-robust anomaly detectors are identified, an autonomous clustering methodology is developed to divide an image into homogeneous background materials, and competing multivariate outlier detection methods are evaluated. To arrive at a final detection algorithm, robust parameter design methods are employed to determine parameter settings that achieve good detection performance over a range of hyperspectral images and targets. The final anomaly detection algorithm is tested against existing local and global anomaly detectors, and is shown to achieve superior detection accuracy when applied to a diverse set of hyperspectral images. The proposed signature matching methodology employs image-based atmospheric correction techniques in an automated process to transform a target reflectance signature library into a set of image signatures. This set of signatures is combined with an existing linear filter to form a target detector that is shown to perform as well or better relative to detectors that rely on complicated, information-intensive atmospheric correction schemes. The performance of the proposed methodology is assessed using a range of target materials in both woodland and desert hyperspectral scenes.
Author:
Publisher:
ISBN:
Category : Algorithms
Languages : en
Pages : 71
Get Book Here
Book Description
The motivation of this work is to investigate the use of data clustering to improve our ability to detect targets within hyperspectral images. Target detection algorithms operate by identifying locations that are likely to contain a target when compared with the background. We propose a new clustering-based target detection method that allows multiple background models to be used. This new method pairs a clustering algorithm with an array of spectral matched filters. We then analyze the performance of various clustering algorithms when used with this method to detect targets in aerial hyperspectral images. We evaluate the performance of our clustered target detector on several aerial hyperspectral images when using clusters generated by several popular algorithms, specifically k-means, spectral clustering, Gaussian mixture models, and two variants of subspace clustering. We show empirically that our tuned algorithm outperforms all others when used for this task, outpacing the traditional Gaussian mixture model with a pAUC score of 0.219 for the same case above, thereby offering over a 14-fold improvement in performance. We offer several hypotheses to explain these results. We then discuss some of the features, most notably the versatility provided by the regularizer, that make the tuned LapGMM algorithm well suited for this application. Considering future work, we propose a number of potential applications for our tuned LapGMM algorithm, as well as several potential improvements or modifications to the clustered target detector that may be worth further investigation.
Author: Christopher Joseph Morman
Publisher:
ISBN:
Category : Multispectral imaging
Languages : en
Pages : 155
Get Book Here
Book Description
Hyperspectral remote sensing has become a popular topic of research due to the numerous applications stemming from the high dimensionality of the recorded spectral data. From the design perspective, hyperspectral sensors are generally more complex than standard color or infrared imaging systems because there are more optical components in the system. The quality of each of these components directly affects the target detection performance of the system. In addition to the integrity of optical components, target detection performance is also affected by signal variations due to sensor noise. This research addresses the design of an end-to-end hyperspectral imaging system performance model that incorporates the optical design of the system as well as the stochastic nature of data collected by electronic remote sensing. A system transmission model is presented that calculates the camera signal as a function of input radiance and accounts for each individual optical element in the imaging system. This model can be used to analyze the performance sensitivities of a specific component for a variety of target detection scenarios. The accuracy of the system transmission model is assessed using calibrated hyperspectral data. In addition to the system transmission model, a realistic statistical data model is proposed. Many data models currently account for sensor noise with an additive, stationary variance. This research expands upon this by implementing an additive, signal-dependent sensor noise model that more accurately represents the true phenomena driving the sensor noise. The same data set is used to test target detection performance using the signal-dependent noise model. The results are analyzed to investigate the possible benefits of using the proposed noise model. The data used for this research was collected at Wright Patterson Air Force Base 25-26 June 2014. The scene consists of a grassy background with eight painted wooden panel targets. Data collections took place at different times of day in order to capture varying solar angles and illumination levels. Additionally, data was collected with varying exposure times in an effort to observe performance effects due to varying signal-to-noise ratios. Conclusions about the performance of the system transmission and data modeling techniques are framed within the context of collection time and exposure time.
Author: Saurabh Prasad
Publisher: Springer Nature
ISBN: 3030386171
Category : Computers
Languages : en
Pages : 464
Get Book Here
Book Description
This book reviews the state of the art in algorithmic approaches addressing the practical challenges that arise with hyperspectral image analysis tasks, with a focus on emerging trends in machine learning and image processing/understanding. It presents advances in deep learning, multiple instance learning, sparse representation based learning, low-dimensional manifold models, anomalous change detection, target recognition, sensor fusion and super-resolution for robust multispectral and hyperspectral image understanding. It presents research from leading international experts who have made foundational contributions in these areas. The book covers a diverse array of applications of multispectral/hyperspectral imagery in the context of these algorithms, including remote sensing, face recognition and biomedicine. This book would be particularly beneficial to graduate students and researchers who are taking advanced courses in (or are working in) the areas of image analysis, machine learning and remote sensing with multi-channel optical imagery. Researchers and professionals in academia and industry working in areas such as electrical engineering, civil and environmental engineering, geosciences and biomedical image processing, who work with multi-channel optical data will find this book useful.
Author: Jason T. Casey
Publisher:
ISBN:
Category : Computer algorithms
Languages : en
Pages : 138
Get Book Here
Book Description
"Line scanning hyperspectral imaging systems are capable of capturing accurate spatial and spectral information about a scene. These data can be useful for detecting sub-pixel targets. Such systems, however, may be limited by certain key characteristics in their design. Systems employing multiple spectrometers, or that collect data from multiple focal planes may suffer an inherent misregistration between sets of collected spectral bands. In order to utilize the full spectrum for target detection purposes, the sets of bands must be registered to each other as precisely as possible. Perfect registration is not possible, due to both the sensor design, and variation in sensor orientation during data acquisition. The issue can cause degradation in the performance of various target detection algorithms. An analysis of algorithms is necessary to determine which perform well when working with misregistered data. In addition, new algorithms may need to be developed which are more robust in these conditions. The work set forth in this thesis will improve the registration between spectral bands in a line scanning hyperspectral sensor by using a geometric model of the sensor along with aircraft orientation parameters to pair sets of image pixels based on their ground locations. Synthetic scenes were created and band-to-band misregistration was induced between the VIS and NIR spectral channels to test the performance of various hyperspectral target detection algorithms when applied to misregistered hyperspectral data. The results for this case studied show geometric algorithms perform well using only the VIS portion of the EM spectrum, and do not always benefit from the addition of NIR bands, even for small amounts of misregistration. Stochastic algorithms appear to be more robust than geometric algorithms for datasets with band-to-band misregistration. The stochastic algorithms tested often benefit from the addition of NIR bands, even for large amounts of misregistration."--Abstract.
Author: Chein-I Chang
Publisher: Springer Science & Business Media
ISBN: 1441991700
Category : Computers
Languages : en
Pages : 372
Get Book Here
Book Description
Hyperspectral Imaging: Techniques for Spectral Detection and Classification is an outgrowth of the research conducted over the years in the Remote Sensing Signal and Image Processing Laboratory (RSSIPL) at the University of Maryland, Baltimore County. It explores applications of statistical signal processing to hyperspectral imaging and further develops non-literal (spectral) techniques for subpixel detection and mixed pixel classification. This text is the first of its kind on the topic and can be considered a recipe book offering various techniques for hyperspectral data exploitation. In particular, some known techniques, such as OSP (Orthogonal Subspace Projection) and CEM (Constrained Energy Minimization) that were previously developed in the RSSIPL, are discussed in great detail. This book is self-contained and can serve as a valuable and useful reference for researchers in academia and practitioners in government and industry.
